ComfyUI Pixaroma

Nodes and canvas tools that make ComfyUI easier to use.

Load, resize, preview and save images. Write and reuse prompts.
Switch parts of a workflow on and off, and explain it with notes.

📖 Need help? Start here

workflows.pixaroma.com/help has short videos with a voice that show every step, the same steps written under each video, questions and answers you can search, and a page for every node.

| I want to... | Open this | |---|---| | Install ComfyUI with the Pixaroma nodes (Windows) | Install with Easy-Install | | Add the Pixaroma nodes to the ComfyUI I have | Install with git | | Update the Pixaroma nodes | How to update | | Fix empty nodes or missing buttons after an update | The 5 second fix | | Fix red "missing nodes" in a workflow | Missing nodes | | Find a model file a workflow needs | Missing models | | See which version I have | Check your version | | Read what a node does and how to use it | Pixaroma nodes | | Ask a question or report a bug | Ask for help |

Inside ComfyUI: click the button in the top toolbar (or press Alt+H). It opens help for every Pixaroma node, with a search that reads the whole text, so you can type the problem ("buttons missing") instead of a node name.

Questions, bugs and ideas all go to #pixaroma-nodes on Discord (join the server first, it is free). Issues and pull requests on GitLab and GitHub are closed, so everything lands in one place and gets an answer.

📥 Install

ComfyUI Easy-Install (Windows): the Pixaroma nodes are already included. Watch how to install it.

With git (portable ComfyUI or your own install): open your ComfyUI/custom_nodes folder, type cmd in the address bar, and run:

git clone https://gitlab.com/pixaroma/comfyui-pixaroma.git

Then restart ComfyUI and press Ctrl+Shift+R in the browser.

ComfyUI Manager: searching for Pixaroma works, but the Manager only offers the old version 1.4.144. The Comfy Registry marked the newer versions as "flagged" with no reason given. The nodes are safe. To get the newest version, use Easy-Install or git, or switch Pixaroma to nightly in the Manager. Read more.

🔄 Update

Easy-Install: close ComfyUI, open the EZi Launcher and press Easy-Install under UPDATE. Video
git: open the ComfyUI-Pixaroma folder inside custom_nodes, type cmd in the address bar, and run git pull. Video
After every update: restart ComfyUI, then press Ctrl+Shift+R once in the browser (Cmd+Shift+R on a Mac). Without it, the browser can keep showing the old version of the nodes. Why
Which version do I have? Click the button in the top toolbar: your version is in the bottom right corner of the Help window. Or add the Version Check Pixaroma node.

🔍 Find the nodes

Double-click an empty spot on the canvas and type Pixaroma: every Pixaroma node shows up. You can also open the Node Library on the left and the 👑 Pixaroma folder, where the nodes are sorted into the groups below.

🧩 All nodes at a glance

The groups are in order of how much they are used in real workflows, and inside each group the most used node comes first. Click a name for its full help on the website: what it does, how to use it, and every input and setting.

🖼️ Image

| Node | What it does | |---|---| | Preview Image Pixaroma | Shows your images on the node, one or a whole batch, with Save, Copy and Open buttons. | | Load Image Mini Pixaroma | A small Load Image: a picker, a preview and two outputs. Resize lives in the gear. | | Image Info Pixaroma | Unpacks Load Image Mini's image_info into the mask, width, height and file name. | | Image Compare Pixaroma | Compare two images with a slider, side by side, or see exactly what changed. | | Load Images from Folder Pixaroma | Runs your workflow once for each picture in any folder you choose. | | Save Image Pixaroma | Saves to any folder, with a live preview of the exact file name. | | Save Mp4 Pixaroma | Turns frames and sound into an mp4 and plays it on the node. | | Remove Background Pixaroma | Cuts out the subject with a BiRefNet model: cutout, mask and inverted mask in one node. | | Load Image Pixaroma | Load Image with resize built in: max megapixels, longest side, fit inside, crop to fill and more. | | Image Grid Pixaroma | Puts every picture of a Run in one grid image, with each file name under its picture. | | Save 3D Pixaroma | Shows a 3D model, stands it on the ground, and saves it as OBJ, GLB or STL. | | Load Video Pixaroma | Loads a video as frames, sound and info, and plays it on the node. | | Pause Image Pixaroma | Stops the Run so you can look at the picture before the slow part runs. | | Load 3D Pixaroma | Loads a 3D model and gives you the model and a picture of it. | | Load Video Frame Pixaroma | Picks one exact frame from a video to use as an image. | | First Last Frame Pixaroma | Takes the first and the last frame of a video as two images. | | Save Video Pixaroma | Saves an mp4 to any folder with your own file name, and plays it on the node. |

💬 Prompt and Text

| Node | What it does | |---|---| | Prompt Pixaroma | A prompt box with saved shortcuts (@tags), random picks and lists. | | Show Text Pixaroma | Shows any text or value going through a wire, in a box you can copy from. | | Text Pixaroma | A big text box with one text output. Type once, wire it into many nodes. | | Text Join Two, Three and Four Pixaroma | Joins two, three or four pieces of text into one. Type or wire each one. | | AI Prompt Pixaroma | Runs a language model on your own PC with an instruction you save on the node. | | Prompt Each Pixaroma | Many prompts, one Run, one picture each. | | Prompt Multi Pixaroma | Queues one Run per prompt, or sends all the prompts out as a list. | | Prompt From List Pixaroma | Picks one prompt by number from a Prompt Multi list. | | Prompt Pack Pixaroma | Paste a block of prompts and queue one Run per prompt. | | Save Text Pixaroma | Collects the text of every Run into one list and a .txt file. | | Prompt Reader Pixaroma | Reads the prompt saved inside a ComfyUI, Automatic1111 or Civitai picture. | | Prompt Stack Pixaroma | Builds one prompt from rows you switch on and off. | | Video Prompt Pixaroma | Writes a MiniMax H3 video prompt on your own PC. | | Find and Replace Pixaroma | Changes words in a text on its way through a wire. | | Load Texts from Folder Pixaroma | Sends the .txt files of a folder through the workflow one by one. | | Pause Text Pixaroma | Stops the Run so you can read and fix a text before the rest runs. | | Caption Review Pixaroma | Every picture of a folder with its caption under it, to read and fix a training set. | | Music Prompt Pixaroma | One idea in, a caption and lyrics out for MiniMax Music 3. | | Layout Boxes Pixaroma | Draw boxes where things go and write what goes in each, for FLUX.3 layout prompts. |

📝 Notes and Overlay

| Node | What it does | |---|---| | Label Pixaroma | A text caption on the canvas to name parts of your workflow. No wires, it never runs. | | Note Pixaroma | A note with headings, lists, tables, icons, and Download or YouTube buttons. | | Info Pixaroma | A small button that opens a note in a reading window, so big notes do not fill the canvas. | | Text Watermark Pixaroma | Stamps text in the same spot on every image of a batch, all set up on the node. | | Text Overlay Pixaroma | Puts styled text on an image, with a full screen editor to place it. |

🔢 Values

| Node | What it does | |---|---| | Seed Pixaroma | A seed for your samplers: Random or Fixed, and one click to keep a good result. | | Sizes Pixaroma | Your own list of favourite sizes, with a Portrait or Landscape flip. | | Resolution Pixaroma | Pick a ratio (1:1, 16:9 and more) and get a width and height that suit AI models. | | Dropdown Pixaroma | A dropdown you fill in yourself: pick a short name, send a long value. | | Duration Pixaroma | Type a length in seconds, get the frame count your video model wants. | | Number Pixaroma | One number field that sends the same value as a whole number and as a decimal. | | Control Panel Pixaroma | Sliders, switches and dropdowns for settings all over your workflow, in one node. | | Portrait Landscape Pixaroma | Flips a width and height between tall and wide with one click. | | WH Pixaroma | Two number fields: width and height. | | Number Pick Pixaroma | One number, picked from a row of buttons you choose yourself. |

🔀 Logic and Flow

| Node | What it does | |---|---| | Run Timer Pixaroma | Times the whole Run and plays a chime when it is done. No wires needed. | | Free VRAM Pixaroma | Frees the graphics card's memory at the exact point you choose. | | Loop Start and Loop End Pixaroma | Repeats part of a workflow a set number of times. | | Combine Pixaroma | Joins two inputs into one batch: images, frames, latents, numbers or text. | | Switch Pixaroma | Many inputs, one output: click a row to choose which one goes through. | | Mute Switch Pixaroma | Switches to turn whole parts of a workflow on and off. | | Monitor Pixaroma | A live readout of how hard your computer is working. | | XY Plot Pixaroma | Compares settings side by side in one labeled grid, with no extra wiring. | | Set and Get Pixaroma | Sends a value across the canvas with no cable: Set stores it under a name, Get reads it. | | Notify Pixaroma | Plays a sound when the Run reaches this node, and times how long it took to get there. | | Switch WH Pixaroma | Chooses between two width and height pairs with one click. | | Switch Source Pixaroma | Flips a whole set of wires between setup A and setup B. | | Group Switch Pixaroma | On and off switches for your groups, in one small panel. | | Text Match Pixaroma | Does a text contain one of your words? Yes or no, plus gates that let things through only on a match. | | Version Check Pixaroma | Shows your Pixaroma, ComfyUI and frontend versions, with a Copy button. | | Run Log Pixaroma | Keeps the last 10 Run times of this workflow on the canvas. |

✂️ Resize and Crop

| Node | What it does | |---|---| | Resize Crop Pixaroma | Makes an image an exact width and height by cropping, never stretching. | | Longest Side Pixaroma | Makes a picture's longest side the size you click, and can crop it to a shape. | | Image Crop Pixaroma | Draw the crop box on the image, no pixel numbers to type. | | Image Uncrop Pixaroma | Puts an edited crop back onto the original, in the exact spot it came from. | | Inpaint Crop Pixaroma | Paint a mask and get a crop sized for the model, ready to inpaint. | | Inpaint Stitch Pixaroma | Puts the inpainted crop back onto the original so the seam does not show. | | Outpaint Pixaroma | Adds space around an image so a model can paint new scenery there. | | Outpaint Stitch Pixaroma | Puts your sharp original back onto the outpaint result, keeping only the new area. | | Image Resize Pixaroma | Resizes an image and its mask, with many modes including Pad. |

🧰 Utility

| Node | What it does | |---|---| | LoRA Loader Pixaroma | Many LoRAs in one node, with trigger words read from each file. | | Krea LoRA Converter Pixaroma | Makes a Krea 2 LoRA trained on fal.ai load in ComfyUI. | | Krea2 Enhancer Pixaroma | Makes a Krea 2 model follow your prompt more closely. |

🎨 Editors

| Node | What it does | |---|---| | Paint Pixaroma | A painting editor with layers, brushes and a smudge tool. | | 3D Builder Pixaroma | Build, light and render a 3D scene inside ComfyUI. | | Image Composer Pixaroma | Arrange, blend and adjust many images on one canvas, in layers. | | Sketch Pixaroma | Mark what to change on a picture and write the change beside it, for edit models. | | AudioReact Pixaroma | Turns a still image into a video that moves with the music. |

🎵 Audio

| Node | What it does | |---|---| | Load Audio Pixaroma | Shows a sound file's shape so you can take just the part you want. | | H3 Audio Sync Pixaroma | Makes a MiniMax H3 video sing your song instead of one it makes up. | | H3 Timeline Pixaroma | Tells one long MiniMax H3 video as scenes, each with its own words and, if needed, a picture. | | H3 Long Video Pixaroma | Makes one MiniMax H3 video longer than 15 seconds, in one run with no cuts. |

🖱️ Canvas tools (no node needed)

| Tool | What it does | |---|---| | Workflows panel | in the top toolbar (or Alt+W): find, open and organise your workflow files, with a picture of each. | | Help window | in the top toolbar (or Alt+H): help for every node, with search. | | Align | Align in the top toolbar: nodes snap into line as you drag them, with a guide line. | | Pixaroma Group | A group box with Run, Mute, Bypass and Fold buttons in its header. | | Node Colors | Right-click a node or group to recolour it, with favourites and copy and paste. | | Adaptive node titles | Title text turns white or dark by itself, so it stays readable on any colour. | | Connection FX | Slots that fit glow while you drag a wire. Off until you turn it on. | | Run button effects | Optional effects on ComfyUI's Run button. | | Your own fonts | Put .ttf or .otf files in ComfyUI/models/fonts and Text Overlay and Text Watermark can use them. |

✂️ Background removal models

Remove Background Pixaroma and the AI background buttons in Image Composer and Paint use a BiRefNet model. Put one of these files in ComfyUI/models/background_removal/:

| Model | File name | VRAM | Best for | |---|---|---|---| | Standard | birefnet.safetensors (424 MB) | 4 to 6 GB | Objects, products, logos. Fast. Download | | High Resolution | birefnet-hr.safetensors (444 MB) | 8 GB+ | Big images with fine edges. Download | | Matting | birefnet-matting.safetensors (444 MB) | 8 GB+ | Hair, fur, lace, soft fabric. Download |

Rename the HR and Matting files after you download them. They arrive as model.safetensors; rename them exactly as in the table, because the name tells Pixaroma which size to work at. FAQ: which model · FAQ: HR or Matting not working

The two editors can also use rembg (pip install rembg, then restart ComfyUI). The node does not need it. FAQ: do I need rembg?

📺 Learn more

Pixaroma on YouTube: a free ComfyUI course, with a new episode almost every week.
workflows.pixaroma.com: the workflows from every episode, free to download.
Help page: install, update, and fixes for the usual problems.

🛠 Changelog

October 23, 2026 · v1.4.230

NEW: Krea2 Enhancer Pixaroma. Makes a Krea 2 model follow your prompt more closely. It gives the same pictures as the Krea2T Enhancer, which its author removed, so our Krea 2 workflows work again.
H3 Timeline: each scene can start with a clean cut, Keep holds the scenes you like while you redo the others, and a shorter Window fits bigger sizes in memory.
Spell check: Pixaroma text boxes follow ComfyUI's own spell check setting.
Run button effects: the arrow next to Run turns orange too on newer ComfyUI, and Run stays orange after Run (Instant).

October 22, 2026 · v1.4.229

NEW: H3 Timeline Pixaroma. Tell one long MiniMax H3 video as scenes, each with its own length, words and, if needed, a picture: a new look, or a person who walks in. One run, no cuts, tested up to 60 seconds.
NEW: H3 Long Video Pixaroma. One MiniMax H3 video longer than 15 seconds, made in one run with no cuts.
Text size: the text boxes of the text and prompt nodes follow ComfyUI's text size setting, and each node can have its own size from the right-click menu.

October 21, 2026 · v1.4.227-v1.4.228

AI Prompt has two MOSS-Image presets: one turns a short idea into a detailed caption, the other writes the caption that makes a picture again.
AI Prompt, Music Prompt and Video Prompt work on more graphics cards: when ComfyUI's fast text writing stops with "Unsupported Flash Attention tensor strides", they write the text the regular way instead.

October 20, 2026 · v1.4.225-v1.4.226

Load Video loads phone videos the right way up: a clip filmed upright on a phone no longer comes out turned sideways and cropped.
Drop a video file on Load Video to load it, like a picture on Load Image. It used to make a second video node instead.
Group Switch keeps the size you drag it to, through switch flips, tab changes and reopening the workflow.
Sizes and Resolution help explains a twice-as-big picture: some other packs' samplers skip the step that fits the latent to Flux 2 and Qwen Image 2.1.

October 18, 2026 · v1.4.223-v1.4.224

Civitai info keeps the seed in Flux 2 and FLUX 3 workflows: a seed, steps or CFG wired into the sampler's side nodes now reach Civitai.
AI Prompt, Music Prompt and Video Prompt get a Faster writing (MTP) switch in the gear. Turn it off if a Qwen3.5 model crashes with a CUDA error.
Workflows panel sorts numbers as numbers: Sort by Name puts 2 before 10 and keeps each folder's workflows together, and the Sort button always names the order shown.

October 17, 2026 · v1.4.221-v1.4.222

Resolution and Sizes get a batch box beside the latent output: one Run makes 1 to 64 pictures.
Control Panel dropdowns show long model names on a wider node, instead of always cutting them after a few words.

October 16, 2026 · v1.4.216-v1.4.220

New node, Layout Boxes Pixaroma: draw where things go for FLUX.3 layout prompts, in Classic and Nodes 2.0.
Switch, Mute Switch, Text Join and Outpaint Stitch keep their wired rows in Nodes 2.0 on ComfyUI 0.40 and newer.
AI Prompt can send the formula as a system message, plus a FLUX 3 prompt upsampler preset.
Control Panel dropdowns take the new model list when you press R, also through Set and Get.
Open buttons no longer say "Popup blocked" on Save Image, Preview Image and Pause Image.
Find and Replace, Prompt Multi and Prompt Stack no longer grow or open as modified.
Dialogs stay open when a drag ends outside them; dragged panels no longer stick to the mouse.
Text Join and XY Plot text boxes get Copy and Paste on right-click; Pixaroma Group draws links through reroutes.

October 15, 2026 · v1.4.215

Video Prompt: Reset, Import and the seed box work on ComfyUI Desktop, and Ctrl+Z in its editor no longer undoes the workflow.
Copy (Clipspace) from Prompt Reader, Save Image or Save Video no longer breaks the next paste.
Scrub bars stop following the mouse after a release outside the window.
Image Uncrop masks keep their shape; Crop and Composer show subfolder pictures; Ctrl+S works in more fields.

October 13, 2026 · v1.4.213-v1.4.214

Load Images from Folder: a new folder starts with every picture picked, so Browse then Run does the whole folder. Use Pick images to choose fewer.
Sizes and Resolution: a new latent output for the KSampler, the same as Empty Latent Image, so you can skip that node. Resolution's spacing now matches Sizes.
Ctrl+S and Ctrl+Enter now work while typing in node fields and settings panels across the pack (they opened the browser's Save page or did nothing).
Resolution: a shared workflow can no longer run code through a saved size, and Custom Ratio keeps your ratio at the size limits.
Control Panel: a Random seed stays Random when you open its number and change nothing.

October 12, 2026 · v1.4.210-v1.4.212

Preview Image and Image Compare: a Run that ends while another workflow tab is open now shows up when you switch back, and never lands on the other tab.
The same fix for every other node that shows a result (Show Text, Pause Text, Save Text, Save Image, AI Prompt, Notify and 20 more): the other tab's nodes are never touched.
AudioReact: the editor preview turns the same way as the saved video for Swirl, Rotate Pulse, Drift, Slit Scan, Wave and Tilt.
A new, shorter README: every node in one line, grouped in order of use, each linking to its own help page on the website.
Help window: Load Audio and H3 Audio Sync now sit in their own Audio group with its music icon, not last with a box icon.

October 11, 2026 · v1.4.209

Load Images from Folder: clearer help for Skip pictures already done. Use it to do a big folder in parts. A stopped Run has saved nothing yet.

October 10, 2026 · v1.4.208

Load Images from Folder: when every picture is already done, the Run now stops quietly with a short notice instead of a red error box.

October 9, 2026 · v1.4.207

Help window: a Latest version button beside your version, opening workflows.pixaroma.com/latest with the newest version and what is new.
Node titles stay readable in the light theme, and in Nodes 2.0 the Show advanced inputs text and nodes inside subgraphs follow the node colour too.
Load Images from Folder: a folder problem shows on the Pick button, and the folder not approved message reads as clear steps.

October 8, 2026 · v1.4.203-v1.4.206

AI Prompt and Music Prompt no longer crash ComfyUI on a list (a folder or Prompt Each): each item gets its own text.
Save Text: new "One file per entry" switch and name input, so every picture gets its own .txt with the same name, ready as training captions.
Load Images from Folder: new path output, and Prompt Reader reads it, so one Run reads the prompts saved in every picture of a folder.
NEW: Image Grid, every picture of a folder Run in one grid, with names and an optional before row.
NEW: Text Match, a yes or no from your words, with gates that filter a whole folder with an AI.
NEW: Load Texts from Folder and Caption Review: read a folder of .txt captions, and fix them in one window.
Load Images from Folder can skip pictures already done, so a big folder carries on where it stopped.
Fixed: right-click a paused video for its menu again; Load Video's sound follows Skip first frames; Image Grid cells fit wide pictures.

October 6, 2026 · v1.4.199-v1.4.202

Save Image and Preview Image: Civitai shows your prompt, settings and model again with "Add Civitai generation info" on. Save images again before uploading.
Civitai info also reads prompts built with String Concatenate and LoRAs from Power Lora Loader (rgthree), and keeps LoRAs with non-Latin names.
Prompt Reader now reads the prompt from JPG and WebP images, including Civitai downloads and Save Image's own JPG and WebP files.
Save Mp4 and Save Video: new "Add Civitai generation info" switch, so Civitai shows the prompt and settings of your videos. Values from Dropdown Pixaroma are read too.
Prompt Reader no longer shows the negative prompt as the prompt when an image has none, and Civitai info keeps the space between joined texts.

October 5, 2026 · v1.4.197-v1.4.198

Text Overlay and Text Watermark no longer scramble pictures from Qwen Image 2.1 and Ming Image; transparent pictures stay transparent, with the text on top.
Control Panel: the Show options list fits long names and has a filter box, and All or None apply to the matching names.
Monitor's help explains why its VRAM number is lower than Crystools or Task Manager.
Newer ComfyUI: workflows no longer show as changed right after opening them with Text Join, Sliders, Dropdown, Switch, Outpaint Stitch or Load 3D nodes.
Newer ComfyUI, Nodes 2.0: Ctrl+Z no longer gets stuck after adding Paint, 3D Builder, Image Composer, Crop, Text Overlay, Notify or 9 other nodes.

October 4, 2026 · v1.4.193-v1.4.196

Pixaroma loads even when a launcher saves ComfyUI's console output to a file; its startup message could stop every node from loading.
Ctrl+Z no longer gets stuck after you add one of 14 nodes, including Paint, 3D Builder, Image Composer, Prompt and Inpaint Crop.
Seed no longer marks a workflow as changed on open, and Ctrl+Z works with it in Nodes 2.0.
Nodes 2.0: Run Timer, Monitor, Info and Label keep their size on open and Ctrl+Z, so undo works again.
The button colour you pick reaches every part of Image Resize and the image loaders, lists and gallery included.
Nodes 2.0: Align snaps collapsed and title-less nodes to the edges you see.
Mute Switch shows its "out" label right after switching to Nodes 2.0.
Workflows panel: Alt+W closes it even with a question open, and a right-click no longer cancels it.
Help and node descriptions call on/off controls switches and buttons.

October 3, 2026 · v1.4.188-v1.4.192

Save Image: Save now writes a Preview run's pictures to your folder without running again.
Nodes 2.0: the wheel scrolls long prompts when "Scroll the field" is set.
Save Image's Civitai info now includes your prompt from Pixaroma prompt nodes.
Prompt boxes: Ctrl+S saves the workflow, and Replace no longer sends hidden line breaks.
Fixed: Image Compare's Overlay opacity follows the mouse wheel again.
Image Crop: Pixaroma dropdowns for Ratio and Alignment, with full names.
Help search finds Image Resize for "upscale" again, and icon-only buttons have names for screen readers.
Ready for newer ComfyUI: renderer switches and removed rows no longer leave stale data behind.
Load Image and Load Image Mini open at the same height every time, and saved workflows no longer reopen taller.
Workflows panel: an open question box now blocks the panel behind it, and only one can be open.

October 2, 2026 · v1.4.185-v1.4.187

AI Prompt and Music Prompt: saving remembers the preset you edited. Its name is filled in, so pressing Enter updates it. An edited built-in one is offered as "(mine)". Music Prompt now also asks before replacing another set with the same name.
Info: Delete. Right-click a button and choose Delete, or press Delete in its reading window, which asks first. Ctrl+Z brings it back.
Prompt Each: the prompt boxes fit their text again when a workflow opens and when you make the node wider, with no scrollbar.
Fixed: in Nodes 2.0, after Ctrl+Z Pixaroma nodes kept their old controls, so what you typed was lost and Info buttons went blank.

October 1, 2026 · v1.4.183-v1.4.184

NEW: Info Pixaroma. A small button on the canvas that opens a note to read. Pick a title, icon and colour, or start from Read me, Download Models, Run Times and more. Also in the canvas right-click menu, under Add Label.
AI Prompt: two Ming Image presets that turn a short idea into the layout prompt Ming Image 0.1 reads best, written by Qwen3.8 27B or Qwen3.5 9B.
Note: lists no longer leave an empty gap above them, and the Code view no longer adds stray empty lines.
Fixed: in Nodes 2.0, Align moved the selected node when you dragged the sidebar or a floating window, or selected text in another node.
Help: hidden storage fields are no longer listed as settings on the Note, Label and Info pages.

September 30, 2026 · v1.4.182

Resize Crop: sharper when it enlarges a picture. It now uses lanczos both ways, the same as ComfyUI's own resize.
Fixed: a "Pixaroma" badge floating above Label, Run Timer and Monitor after the latest ComfyUI update.

September 29, 2026 · v1.4.177-v1.4.181

Load Image and Load Image Mini: pictures from your own folders (+ Folder in the picker), refreshed from the original at each Run. Load Image also gets a settings gear.
AI Prompt: several pictures at once. Wiring a picture adds a slot for another, up to 8, each seen at its own size.
Paint: "Canvas size from added image". Turn it on under Add Image and every picture you add sets the canvas to its size.
Fixed: Inpaint Crop behind an Image Composer showed the picture going into the Composer, so the mask was painted on the wrong image.
Fixed: Ctrl+V pasted nothing while a Load Image or Load Image Mini was selected, not even copied nodes. Pasting a picture no longer asks Firefox for permission each time.
Save Image: a JPG keeps the prompt when the workflow is too big to fit inside it.
Save Image and Save Video: "Will save as" shows ‹wired name› when the name is only made during the run, instead of a wrong name.

September 28, 2026 · v1.4.176

AI Prompt's "Qwen Image 2.1 - prompt from an idea" preset now writes real selfies, not photos of someone holding up a phone. Load the preset again to get it.

September 27, 2026 · v1.4.170-v1.4.175

New: Sketch Pixaroma. Mark a picture with boxes, circles, arrows or words and a note each; edit models like Flux 2 Klein get the marked picture and a ready prompt.
Fixed: Set and Get Pixaroma broke after refreshing nodes (R) until a page reload. Affected workflows repair themselves, and both are back in the Add Node menu.
Fixed: all Pixaroma nodes showed up empty when ComfyUI runs under a web address folder (a reverse proxy like yoursite.com/comfyui/).
Plus fixes: Inter Bold now comes out bold in Text Overlay and Watermark; Align no longer shrinks a node below its content; XY Plot, Load Images from Folder, Image Resize and Notify keep their size; Civitai info no longer writes the positive prompt as the negative.

September 26, 2026 · v1.4.163-v1.4.169

Run Timer, Run Log and Monitor Pixaroma no longer slow your renders. Run Timer with 2 or 3 decimals had cost 6-9%; its final time stays exact.
Faster renders in Nodes 2.0 with Compare, Preview Image, Outpaint, Load 3D, Save 3D or a video node (Load Video, Load Video Frame, Save Mp4, Save Video) on screen. Their previews show a still picture while nothing moves. Load Audio, Image Resize and Load Image Mini do the same.
Fixed: copied Run Timers, Monitors and Image Composers kept working unseen, even after the node was deleted.
17 nodes now follow switching Nodes 2.0 on or off without a reload, instead of showing empty, doubled or missing parts, or dots that would not take a wire.
Fixed: nodes left inside a subgraph kept acting after a tab switch, disturbing another workflow's nodes or downloading a picture again after every run.

September 24, 2026 · v1.4.162

New: four Qwen Image 2.1 presets for AI Prompt Pixaroma: a prompt from an idea, a transparent picture, a prompt from an image, and an edit instruction.
Fixed: Inpaint Stitch Pixaroma pasted a grey patch when the model returned a picture with transparency, as Qwen Image 2.1 does. It pastes clean colours now.

September 22, 2026 · v1.4.159-v1.4.161

Fixed: Music Prompt Pixaroma closed ComfyUI part way through a run, on ComfyUI 0.37.0. It writes the caption and the lyrics again.
Fixed: in Prompt Pixaroma the cursor could sit away from the words you typed, further out the longer the prompt. The box now follows your theme's font exactly.
A file name can end with _ again. Type name_%counter%_ in Save Image Pixaroma and you get name_001_.png, so you can add your own words after the number.

September 21, 2026 · v1.4.158

Fixed: the prompt and text nodes used the wrong look for your theme: white on a dark one, dark on a light one. A new Node look setting can force either.
Fixed: Load Audio Pixaroma's file list ran off the screen. It scrolls now.
Group Switch Pixaroma no longer slows things down while on screen.
Inpaint Crop Pixaroma follows a Switch, and stops showing a picture that is no longer wired in.

September 18, 2026 · v1.4.154-v1.4.157

New node: Number Pick Pixaroma. One number, picked from buttons you choose yourself. It sends a whole number or a decimal to suit whatever you wire it to.
Volume on the video players. A speaker on Save Mp4 and Save Video Pixaroma. The level you set is remembered by every Pixaroma video node.
Save Mp4 Pixaroma takes a whole video, not only frames, and it now shows the video size on the node, as does Load Video Pixaroma.
Group Switch Pixaroma: an arrow on each row jumps the canvas to that group.
Fixed: Inpaint Crop Pixaroma was blank when a Switch or a Reroute sat between it and the picture.
Fixed: Prompt Reader Pixaroma could open a little short in a big workflow.

September 17, 2026 · v1.4.152-v1.4.153

Fixed: middle-mouse drag did not move the canvas over a node's picture or text box. Now it does, over 3D views too.
Fixed: LoRA Loader forgot your trigger words when a LoRA was moved to another folder. Pick the moved file again and they come back.
Fixed: Ctrl+Up / Ctrl+Down did not change a word's weight in Pixaroma prompt boxes.
Fixed: in ComfyUI's light theme, the prompt and text nodes had invisible buttons and black text boxes.
Fixed: Save 3D views could stay on "Loading ..." until a page refresh, after a Load 3D node had been open.
Fixed: a Save 3D node could show the model from another workflow. Each workflow now keeps its own preview.
If the browser stops drawing 3D views, Load 3D and Save 3D now say so and try again on their own, instead of staying dark.
The info line under Save 3D's view wraps to two lines, so nothing is cut off.

September 16, 2026 · v1.4.151

NEW: Save 3D Pixaroma. See a 3D model on the node, check which way it faces, stand it on the ground, and save it as OBJ, GLB or STL.
Preview Image Pixaroma no longer makes everything after it run again. Change a late setting and only the steps after it are redone, so 3D workflows are much quicker to tweak.

September 14, 2026 · v1.4.150

Load 3D Pixaroma now gives out its width and height, so your empty latent can match the picture.
Load 3D can take its size from Sizes Pixaroma. Wire the two together and the 3D picture follows your size pick, so tall and wide pictures keep the angle of your view.

September 13, 2026 · v1.4.149

NEW: Load 3D Pixaroma. Load a GLB, OBJ, FBX, STL or PLY model, turn it to any view, and get the model, its picture and a mask.

September 11, 2026 · v1.4.145-v1.4.148

Fixed: bypassing a node could break the wire below it. Show Text, Switch, Notify and Free VRAM stopped with "missing input" or passed the wrong value.
Fixed: Save Text kept only the first result of a folder run.
Fixed: a picked button colour came back orange after a reload, hover too.
Fixed: opening and closing a workflow asked "Save changes?" with nothing changed.
Fixed: the settings gear on Prompt Each and Music Prompt opened the colour panel.
The video nodes now install ffmpeg with the pack, so they work on a fresh setup. Update through Manager or your launcher, not git on its own.
Remove Background names every folder it searched when a model is missing.
Clearer help for Control Panel's Step and for Mute Switch.

September 7, 2026 · v1.4.144

Fixed: selecting text in Prompt, AI Prompt and Music Prompt showed no highlight. The words were selected and copying worked, but nothing showed, so there was no way to tell.

September 6, 2026 · v1.4.143

Fixed: a 16-bit picture came out almost white, in Inpaint Crop, Image Crop, Image Composer layers and AudioReact. The preview looked right the whole time, so it read as the node being broken.
The download is smaller. Developer tools and notes that never run on your machine are no longer packed into the release.
The settings gear on Outpaint and Prompt is now drawn by the pack, so it looks the same on Windows, Mac and Linux.

September 5, 2026 · v1.4.133-v1.4.142

Fixed: Preview Image Save to Disk - big pictures failed, the filename stopped counting up, and a refused folder picker now saves to Downloads.
Fixed: the Workflows panel now sees folders you linked in from elsewhere, and can rename, move and cover them. Thanks to jab416171.
Fixed: Prompt Reader returned an AI node's instructions instead of the prompt.
Fixed: Load Audio now finds sound files in folders inside input.
Fixed: a Pixaroma group inside another group could stop being editable.
Fixed: text looked doubled while you selected it in AI Prompt, Prompt and Music Prompt.
Fixed: Run Timer grew a little on every open with a font of your own.
Open folder and Reveal now work only on the computer ComfyUI runs on, since that is whose screen the window appears on.
Groundwork for clearing the Registry security warning. Nothing changes in how the nodes behave.
Reports and ideas now go to #pixaroma-nodes on Discord.

August 27, 2026 · v1.4.131-v1.4.132

NEW: Prompt Each Pixaroma. Type a prompt in each row, press Run once, and you get one picture per prompt, all collected in the same Preview node. It renders them one at a time, so a long list is safe on a small graphics card.
Square brackets multiply a row: a [red|blue] car is two prompts, a [red|blue] [car|van] is all four. The counter shows what you will get before you press Run.
Paste a whole list in one go, one prompt per line, and Copy sends it back out the same way with the ON/OFF switches kept.
It works with your tag library. Wire Prompt Pixaroma into its text input and your @tags are filled in first, added to your rows rather than replacing them.
Fixed properly: Prompt Multi and Prompt Stack rows collapsing into thin overlapping bars. The cause was Monitor Pixaroma restyling them, not another node pack, which is why it came and went and a refresh cleared it.
Prompt rows now hold their height if anything else tries to squeeze them.

August 26, 2026 · v1.4.127-v1.4.130

Fixed: Load Audio grew taller by itself. The waveform pushed the node bigger on every redraw, ending at a different size after each reload and dragging the canvas.
Panning and zooming are lighter on big workflows. Label, Monitor and Run Timer no longer tax the canvas.
Fixed: Save Video, Save Image and Save Text would not fill the node, leaving the video small with empty space below.
Fixed: Prompt Multi and Prompt Stack rows could collapse into each other, only alongside certain other node packs.
Fixed: dragging a saved video back onto the canvas rebuilds the workflow again.
Fixed: Image Crop gave every queued job the last crop.
The LoRA picker no longer opens on an empty folder, and Back is easier to see.

August 25, 2026 · v1.4.123-1.4.126

Dropdown Pixaroma now holds up to four values in one entry. Name each output, and one pick sets several wires at once: a sampler and its scheduler.
Fixed: a Dropdown would not connect to a sampler or scheduler by hand. It worked if the wire was already saved, but once unplugged it could not be plugged back.
Each output has its own type. Existing Dropdowns keep working.
New: Save Mp4 and Save Video can fade the sound in. AI video clips often start with a click; 120 removes it.
Fixed: Save Mp4 could skip saving, with no error, when run from an API script.

August 24, 2026 · v1.4.121-1.4.122

NEW: Free VRAM Pixaroma, hands the graphics card's memory back at the point you wire it in, so a second heavy model has room to load. Under Logic & Flow.
NEW: Monitor Pixaroma, a l

... (README truncated for length)

pixaroma/ComfyUI-Pixaroma: Pixaroma ComfyUI Nodes